摘要
传统时频分析方法可以同时在时、频域反映滚动轴承的故障特征,然而故障信号中的噪声干扰会模糊故障特征在时频分布上的清晰表示。稀疏表征方法可以通过稀疏的原子来表达原始信号,其某些类型的原子在时频域上的聚集是其进行时频分析的良好特性。相比于STFT,基于稀疏原子重构的信号时频分布具有更好的分辨率;相比于WVD则避免了交叉干扰项的影响。提出基于稀疏表征时频分析方法的滚动轴承故障特征提取方法,通过仿真及实验验证所述方法相对于传统时频分析方法如STFT、WVD具有更清晰的时频特征提取效果。
The traditional time- frequency method can reflect the fault feature of rolling element bearing simultaneously. However, the noise contained in the fault signal will fuzzy the analysis results of traditional time- frequency methods. The sparse representation method can express the original signal by sparse atoms, and better time- frequency analysis result can be obtained with the gathering property of some atoms in time-frequency domain. The advantage of the sparse representation time-frequency analysis method over short time frequency transform （STFT） is the advantage of better resolution, and it avoids the influence of cross interference item compared with wigner vile distribution （WVD）. The feature extraction method of rolling element bearing fault signal based on sparse representation time-frequency analysis is proposed in the paper, and simulation and experiment analysis results verify that better time-frequency results can be obtained over the STFT and WVD methods.
出处
《机械设计与研究》
CSCD
北大核心
2017年第5期107-109,114,共4页
Machine Design And Research
关键词
稀疏表征
时频分析
滚动轴承
特征提取
sparse representation
time-frequency analysis
rolling element bearing
feature extraction
